Dear experts
Apologies for the length of this, its really just a quick couple of questions to check that i understand the previous messages ive read on this topic.

Ive run a concat ICA across subjects data from a block design task.
Im interested in whats happening in certain RSNs during the periods of my task and so i simply want to get single subject timecourses for certain ICs.
I think i've got these from the 1st stage output of the dual_regression script?

Am i right in thinking that dual_reg does spatial regression of the IC maps against the filtered_func_data to get temporal dynamics for each component and each subject, and then in the 2nd stage uses these as regressors in a standard glm to get stat-maps? and if you want you can then use randomise to look for differences in these maps between subjects?

So can the dual_reg timecourses be thought of as the fluctuations in BOLD signal for the chosen RSNs in each subject?
Why exactly are these more suitable for further analysis than the melodic_mix timecourses?

Finally, a more general (and probably stupid) question. 
Im confused about the 'sign' of IC timecourses. is it possible to be sure about the sign of a melodic IC timecouse wrt another IC?
i understand that areas of red-yellow in the GGMM IC maps mean the timecourse is positively associated with those areas.
but can you be sure that, for example, a high value in an IC timecourse means high BOLD signal in that brain area in that volume? 

or put another way. say the analysed data was from block task of a demanding visual experiment. In IC1 you see the visual areas positively correlated with your task. In IC2 you'd see the DFM which you know is likely to be anti-correlated. Would you expect the visual areas in IC1 to be red and the DFM areas in IC2 to be blue? or would they both be red? If they were both red would the only way to be sure about the sign of the time course to use a further step like dual_regression?

similarly for a concat analysis across subjects, i assume that (for a given IC) all subjects timecourses have the same sign?

any comments and advice much appreciated
many thanks 
steve